OpenAI Commits $10 Billion to Cerebras Chips in Strategic Shift from Nvidia
OpenAI has inked a landmark $10 billion agreement with Cerebras Systems, securing up to 750 megawatts of AI computing capacity through 2028. The deal marks a deliberate diversification play—both for OpenAI's hardware stack and Cerebras' client base, which previously relied on G42 for 87% of revenue.
Cerebras' wafer-scale systems promise latency reductions for ChatGPT while sidestepping Nvidia's supply constraints. Early benchmarks showing efficient operation with OpenAI's open-source models catalyzed the August 2023 negotiations. Deployment begins Q1 2026, scaling to what Cerebras CEO Andrew Feldman calls "the largest high-speed AI cluster ever built."
The chipmaker withdrew IPO paperwork last October but intends to refile with updated financials reflecting this anchor contract. For OpenAI, the MOVE signals both the urgency of securing exascale compute and the industry's scramble for alternatives to CUDA dominance.
